Nine key genomic features of Sarbecovirus

In our previous study [2], key genomic features were identified through the analysis of mutation sites in Sarbecovirus. Firstly, mutations were identified in the Sarbecovirus genomes, and then classified into two categories: small- and large-scale mutations. Small-scale mutations included single nucleotide substitutions (SNSs) and small insertions or deletions (InDels), whereas large-scale mutations included segment substitutions and large InDels involving at least 10 nucleotides. Secondly, by small InDels at six specific sites, Sarbecovirus was divided into two classes: the SARS1 class, which included SARS-CoV (from patients) and SARS-like CoV (from animals), and the SARS2 class, which included SARS-CoV-2 (from patients) and SARS2-like CoV (from animals). Thirdly, large-scale mutations occurring within and between SARS1 and SARS2 classes were detected using 292 Sarbecovirus genomes [2]. Finally, we found that almost all the detected large-scale mutations concentrated in nine genomic regions, identified as nine key genomic features. In addition, the genotypes of genomic features observed in the SARS-CoV-2 genome (GenBank: MN908947) were designated as wild-type.

Here, nine key genomic features (designated RC0-7 and ORF8) are introduced (Fig 1). The first key genomic feature, designated as recombinant region 0 (RC0), was identified as the first crucial clue for the origin tracing of SARS-CoV-2 [2]. Wild-type RC0, encoding the junction FCS of SARS-CoV-2 [5], remains undetected outside of the SARS-CoV-2 genome. Another notable key genomic feature is the ORF8 gene [10], which is the crucial clue for the origin tracing of SARS-CoV; however, it has no special significance for SARS-CoV-2. The other seven key genomic features were designated as RC1-7, which encode 12- to 42-amino-acid(aa) peptides or protein segments. RC1-2. RC3-5, and RC6-7 encode peptides in the NSP3 protein [17], N-terminal domain (NTD) and RBD of the S1 subunit, respectively (Fig 1). In addition, a new genomic feature, designated as RC8, was identified during the present study. RC8 encodes the first approximately 30 aa residues, including the signal peptide of the S protein, and exhibites even higher diversity than RC1-7, as evidenced by the detection of many frame-shift mutations within it. Although the diversity of RC0-8 and ORF8 was validated using new genome data (Materials and Methods), only nine (RC0-7 and ORF8) were used in our data analysis. Notably, a comprehensive study of 4,976,200 SARS-CoV-2 variants [18] revealed that InDels are still highly frequently ocurring in RC3-5 during the COVID-19 pandemic, while RC1-2 and RC6-7 remain relatively conservative, with no highly frequent InDels detected in them.

Our previous study [2] revealed that the peptides or protein segments encoded by RC0-7 are disordered in their secondary structures, and their specific functions remain largely unknown, except for RC6 and RC7. By analyzing the structure data [2], we predicted that the peptides encoded by RC6 and RC7 contain all key aa residues in the interaction of RBD with its receptor angiotensin converting enzyme 2 (ACE2). As almost all the detected large-scale mutations in RBD concentrated in RC6 and RC7, we designated them as key genomic features, rather than the entire RBD. These disordered peptides or protein segments, also referred to as hypervariable regions (HVRs), are typically involved in protein-protein interactions (PPIs). Thus, they may have regulatory rather than essential roles in the CoV life cycle, allowing them to tolerate numerous large-scale mutations. In our previous study [2], we linked the mutations in HVRs, especially the nine key genomic features, to the adaptability of Sarbecovirus for new hosts or even the expansion of their host range [19]. However, it is unclear which specific mutations in RBDs enhance their interactions with host receptors and which specific mutations contribute to immune evasion, as the mechanisms underlying the adaptability remain unknown. In the comprehensive study of 4,976,200 SARS-CoV-2 variants [18], reseachers also identified HVRs encoded by RC1-5, along with additional HVRs. In addition, they proposed that the increased frequency of indels, the non-random distribution of them and their independent co-occurrence in several variants of concern (VOCs) is another mechanism of response to elevated global population immunity [18]. Unfortunately, they did not reveal the underlying mechanisms at the molecular level, probably because they did not investigate the connection between mutations and the jumping transcription of CoVs [8, 9].

Integrating the results from our previous study [2] and the previous study [14], several important findings were obtained, including: (1) CoV HVRs experience highly frequent InDels, compared to other genomic regions; (2) deletions are more prevalent than insertions within HVRs; (3) several notable large deletions (longer than 10 nucleotides) in HVRs of SARS-CoV and SARS-CoV-2 were reported in attenuated strains [20–22]; and (4) each of the nine key genomic features has a few distinct genotypes. According to our model explaining the jumping transcription of CoVs [8, 9], the typical occurrence of jumping transcription relies on two types of transcription regulatory sequences (TRSs) successfully matching: leader TRSs (TRS-Ls) and body TRSs (TRS-Bs). Recombination occurs when the antisense sequences of TRS-Bs anneal or connect to regions containing TRS-like sequences, rather than TRS-Ls in the RNA synthesis process. Although CoV recombination is highly frequent, most of recombination events result in small-scale mutations, defined as the type 1 recombination, while only a few of them result in large-scale mutations, defined as the type 2 recombination. One genotype of a key genomic feature changes into another genotype relying on type 2 recombination, rather than evolves into another genotype through gradual accumulation of type 1 recombination. A large InDel results from a single recombination event, rather than gradual accumulation of small-scale mutations. As a result, the genotypes of the nine key genomic features observed in wild types of SARS-CoV-2 remain unchanged or undergo only minor modifications after a short period of evolution. Consequently, all the nine wild-type features remain detectable in one or several Sarbecovirus populations after the COVID-19 outbreak, particularly in the region where SARS-CoV-2 originated. This hypothesis allows for tracing SARS-CoV-2’s genetic and geographical origins by only detection of the nine wild-type features before piecing together the whole puzzle.

Three close relatives of SARS-CoV-2

The relationship between newly discovered Sarbecovirus strains and SARS-CoV-2 were simply measured by their nt identities at the genome level. Regarding genomic features, most of previous studies focused on the RBD corresponding to RC6-7, while only a few of them examined the junction FCS corresponding to RC0 [5]. RaTG13 was identified as the closest relative to SARS-CoV-2, soon after the COVID-19 outbreak; however, its genomic features were insufficient to fully elucidate the origin of SARS-CoV-2. Two years later, a closer relative BANAL-20-52 was discovered with a higher nt identity with SARS-CoV-2. Furthermore, BANAL-20-52 shares almost identical RBDs and ORF8 with SARS-CoV-2, yet it lacks the junction FCS as all other bat Sarbecovirus strains. According to our proposal, the researchers who discovered BANAL-20-52 examined only three of the nine genomic features (RC6-7 and ORF8). By examining the nine key genomic features (Fig 2), it was found that BANAL-20-52 harbor seven wild-type features (RC2-7 and ORF8). Subsequently, we expanded our examination to include new genome data of Sarbecovirus submitted after the COVID-19 outbreak, by setting the criteria (Materials and Methods). As a result, a total of three strains (Fig 1) were identified as close relatives of SARS-CoV-2, harboring seven wild-type features (Table 1). They are BANAL-20-52 [4], Rp22DB159 [11], and S18CXBatR24 [12], isolated from malayan horseshoe bats (Rhinolophus malayanus) in Laos, least horseshoe bats (Rhinolophus pusillu) in Vietnam, and marshall’s horseshoe bats (Rhinolophus marshalli) in Yunan of China, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Crucial evidence for origin tracing of SARS-CoV-2.

SARS-CoV-2 is the Wuhan-2019 strain used as the reference. BANAL-20-52, S18CXBatR24, Rp22DB159, RaTG13, BANAL-20-103, BANAL-20-236, BANAL-20-116, and BANAL-20-247 are bat Sarbecovirus strains, while MP789 and PCoV_GX-P1E are pangolin bat Sarbecovirus strains. The accession numbers of their genomes were provided in the Materials and Methods. As three of the nine key genomic features in Sarbecovirus (Fig 1), RC1 and RC6-7 encode peptides in the NSP3 protein and RBD of the S1 subunit in the S protein, respectively. The dots indicate aa or nt residues same to the reference. The mutant aa or nt residues are indicated by red color, while the mutant segments (different genotypes) are indicated by green color. Read1 (ID: 14445348.2) and Read2 (ID: 14755068.2) are two reads from the NGS data (SRA: SRR22719882) of the sample containing BANAL-20-236. Read1 and Read2 represent reads which were aligned to wild-type RC1 in SARS-CoV-2 and those which were aligned to mutant RC1 in BANAL-20-236, respectively, as they entirely covered wild-type RC1 and mutant RC1, respectively.

https://doi.org/10.1371/journal.pone.0309557.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Key genomic features examined in Sarbecovirus strains.

https://doi.org/10.1371/journal.pone.0309557.t001

By examining the nine key genomic features, many potential "donor strains" were excluded from close relatives, although a certain of them share high nt identites (>90%) with SARS-CoV-2 at the genome level as the three close relatives. Despite harboring five wild-type features (RC2-5 and ORF8), RaTG13 was still excluded (Table 1), according to our criteria (Materials and Methods). Before the present study, there was no consensus understanding of the relationship between RaTG13 and SARS-CoV-2, mainly because of significant differences observed in their RBDs, equivalent to RC6 and RC7 (Fig 2B). These differences led to a speculation that the emergence of SARS-CoV-2 may have depended on recombination or natural selection to enhance its RBD affinity to human ACE2 [4]. Later, the speculation was challenged by the discovery of BANAL-20-52 and was finally refuted by our identification of the three close relatives which harbor RBDs with similar affinity to human ACE2 as SARS-CoV-2. Several pangolin Sarbecovirus strains had been considered as “donor strains”, also because they harbor RBDs with similar affinity to human ACE2 as SARS-CoV-2. Among them, MP789 shares almost identical RC6 and RC7 with SARS-CoV-2 (Fig 2B). However, MP789 and another strain PCoV_GX-P1E [23], harboring only two (RC6 and RC7) and one (i.g., RC7) wild-type features respectively, were excluded from close relatives. Two notable strains RmYN01 and RmYN02, harboring only 0 and two (RC1 and RC2) wild-type features respectively, were also excluded, although nature insertions in their RC0s had been reported in a previous study [24]. Furthermore, RmYN02 possesses a unique recombinant ORF8 identified in our previous study [2], which may have resulted from a rare recombination event between Sarbecovirus strains of the SAR1 class and those of the SAR2 class. In a broader sense, almost all Sarbecovirus strains can be considered as potential "donor strains" contributing to the genesis of SARS-CoV-2, given their high nt identities with SARS-CoV-2 at the genome level or even in a very small genomic region. Identification of all the recombination events involving these "donor strains" in the evolutionary history could not provide significant clues for the origin tracing of SARS-CoV-2. Instead, comparing the genomes of the three or more close relatives may provide insights into the latest recombination events, which made key or even crucial contributions to the genesis of SARS-CoV-2.

While Rp22DB159 and S18CXBatR24 exhibit slightly lower nt identities (92.30% and 92.19%,) with SARS-CoV-2, respectively, compared to RaTG13 (96.13%), they possess two additional key genomic features (RC6 and RC7) in wild genotypes (Fig 2B), absent in RaTG13. The three close relatives (BANAL-20-52, Rp22DB159, and S18CXBatR24) habor wild-type NTDs and RBDs. Particularly, the genomic regions encoding the RBDs of the three close relatives have very high nt identities of 93.87% (628/669), 93.87% (628/669) and 93.57% (626/669) with SARS-CoV-2, respectively, which were wild-types. Compared to SARS-CoV-2, there are nine single aa substitution sites in RBDs of the three close relatives (S1 Text), including four common aa substitutions (A372T, Q498H, H519N, and V534I) shared by the three relatives, two aa substitutions (E324Q and F486L) specific to Rp22DB159 and S18CXBatR24, two aa substitutions (R346T and I402V) specific to BANAL-20-52, and a unique aa substitution N501Y present only in S18CXBatR24. Among these single aa substitutions, only F486L, Q498H, and N501Y were detected to be associated with the 17 aa residues of the RBD that interact with human ACE2 [4]. By comparing the RBD-ACE2 complex structures of SARS-CoV-2 with those of the three close relatives (Materials and Methods), it can be observed that F486L only causes a slight decrease in the hydrophobic surface at position 486, which may not significantly impact the interaction between the RBD and human ACE2. The notable Q498H, shared by three close relatives, was reported to enhance the affinity of the SARS-CoV-2 RBD to human ACE2, and also to be involved in the host range expansion of SARS-CoV-2 and SARS2-like CoVs [4]. N501Y, along with N501T are two common mutations at the N501 position present in at least five VOCs, including Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2) and Omicron (B.1.1.529). Detected in approximately 45.7% and 0.00043% of SARS-CoV-2 variants, N501Y and N501T may enhance and reduce the affinity of the SARS-CoV-2 RBD to human ACE2, respectively [25]. According to a comprehensive analysis with validation by experiments [4], BANAL-20-52 possesses a RBD with enhanced affinity to human ACE2, primarily due to Q498H. Therefore, Rp22DB159, which harbors Q498H, possesses a RBD with the similar affinity as BANAL-20-52, while S18CXBatR24 could possess a RBD with even higher affinity, as it harbors both Q498H and N501Y.

The seven wild-type features in the three close relatives preliminarily validated our hypothesis that all key genomic features of SARS-CoV-2 in wild genotypes remain detectable in one or several Sarbecovirus populations in the region where SARS-CoV-2 originated. Unexpectedly, the three close relatives haboring seven wild-type features were identified in samples independently collected from locations distant from each other. Hence, future research should have focused on the search for the two remaining key genomic features (RC0 and RC1) in wild genotypes within the three regions where the three close relatives had been discovered. However, the crucial clue –wild-type RC0 remains undetected outside of the SARS-CoV-2 genome, which is consistent with our previous prediction [5] that it may have been acquired in intermediate hosts or humans. Fortunately, RC1 was identified as the second crucial clue during the present study, as RC1 and its encoded segment exhibited large differences between SARS-CoV-2 and its three close relatives (Fig 2A), compared to the seven key genomic features (RC2-7 and ORF8) that are almost identical between SARS-CoV-2 and its three close relatives. Wild-type RC1 encodes a 42-aa segment (NSP3: 165–206 aa) localized in an HVR of NSP3. This HVR was annotated as spanning 113–206 aa residues of the NSP3 protein in SARS-CoV-2, by analysis of the structures of its neighboring domains (Materials and Methods). The similarities between the 42-aa segment of SARS-CoV-2 and their homologs in the three close relatives are even lower than that of RmYN02, which harbors seven single aa substitutions. The homologs in Rp22DB159 and S18CXBatR24 harbor nine and ten single aa substitutions, respectively, while that in BANAL-20-52 exhibits one single aa substitution and an 8-aa deletion—TIQTIVEV (Fig 2A). Using the 42-aa segment to search for closer homologs, numerous homologs within the realm of animal Sarbecovirus were obtained from NCBI NT and NR databases. Among them, the closest homolog containing six single aa substitutions was identified in Rp22DB167, which had been isolated from the collection site of Rp22DB159. However, the quality of the Rp22DB167 genome is considerate low. All other homologs exhibit more than six single aa substitutions compared to the wild-type 42-aa segment.

The discovery of wild-type RC1 as crucial evidence

Following the identification of RC0, RC1 was identified as the second crucial clue during the present study, its uniqueness underscored its particular significance. Notably, the 8-aa deletion in NSP3 of BANAL-20-52 (corresponding to a 24-nt deletion in RC1) is analogous to the 4-aa deletion (corresponding to a 12-nt deletion in RC0) resulting in loss of the junction FCS in all other SARS-2 like strains [5]. The RC1 of BANAL-20-52 contains only one single nucleotide polymorphism (SNP) and a 24-nt deletion, while those in Rp22DB159 and S18CXBatR24 contain nine and 11 SNPs, respectively (Fig 2A). According to our findings (Described above), the 24-nt deletion in RC1 or the 12-nt deletion in RC0 could have resulted from a single recombination event, rather than gradual accumulation of small-scale mutations. This suggested that the RC1 of BANAL-20-52 had diverged from that of SARS-CoV-2 relatively late, making the detection of wild-type RC1 highly probable in a Sarbecovirus population BANAL-20-52 belonging to. Further research should have focused on the search for wild-type RC1 within the region where BANAL-20-52 had been discovered, using additional genome data, especially of bat individuals or populations. Given the unavailability of additional genomes, we endeavored to search for it using the Next-Generation Sequencing (NGS) data of Sarbecovirus strains collected in the previous study [4], where BANAL-20-52 was reported. We believed that wild-type RC1 could be detected as low-depth contigs from minor genotypes, which might have been misassembled due to the limitations of the short reads produced by NGS technologies. In a sense, each viral genome sequence assembled using the NGS data is a mosiac genome sequence. It is assembled as a consensus representation of the genetic diversity within a virus population, reflecting the predominant genotypes at each position. However, low-depth contigs from minor genotypes, absent in the genome, are usually neglected for analysis. Although all genome sequences of individuals in a viral population can be obtained using advanced sequencing technologies such as PacBio and Nanopore DNA-seq [26], the resulting data typically includes sequences from major genotypes, with the risk of missing the sequences from minor genotypes.

As expected, wild-type RC1 was finally detected, albeit in the sample containing BANAL-20-236, rather than that containing BANAL-20-52. In the NGS data (SRA: SRR22719882) of the sample containing BANAL-20-236, numerous reads longer than 100 bp were aligned to the mutant RC1 in BANAL-20-236, while a few reads were aligned to wild-type RC1 in SARS-CoV-2 (Fig 2C). Among these reads, 56,168 reads entirely covered mutant RC1, while only two reads entirely covered wild-type RC1. The uniqueness of wild-type RC1 ruled out the possibility that its reads resulted from other sources. The reads aligned to wild-type RC1 had not been assembled into the BANAL-20-236 genome, due to their lower abundance compared to those aligned to mutant RC1. At first glance, this suggested that a Sarbecovirus population, BANAL-20-236 belonging to, possessed eight wild-type features (RC1-7 and ORF8). However, based on similarities measured by the seven key genomic features, all five Sarbecovirus strains (Table 1) reported in the previous study [4] can be classified into three populations and they are: (1) the BANAL-20-236 population including at least the BANAL-20-103 and BANAL-20-236 strains; (2) the BANAL-20-116 population including at least the BANAL-20-116 and BANAL-20-247 strains, and (3) the BANAL-20-52 population including at least itself. Among these strains, BANAL-20-52 is the closest relative of SARS-CoV-2, while BANAL-20-236 exhibits the largest differences from SARS-CoV-2 in their genomes. It is more likely that wild-type RC1 was acquried by BANAL-20-52 than BANAL-20-236 for the genesis of SARS-CoV-2. Therefore, wild-type RC1 may also be present in the BANAL-20-52 populaiton as a minor genotype. If so, a potential closest relative could be assembled using the NGS data of BANAL-20-52 or other samples. This closest relative, named BANAL-RC1, harboring eight wild-type features (RC1-7 and ORF8), is even closer to SARS-CoV-2 than BANAL-20-52 which harbors seven wild-type features. It is conceivable that BANAL-RC1 was previously and is still present within this Sarbecovirus population. However, the NGS data of BANAL-20-52 was not available for validation. Nevertheless, wild-type RC1 was detected in a large bat Sarbecovirus population BANAL-20-52 and BANAL-20-236 belonging to, which dispersed over a region with a diameter at least 2.22 km, harboring eight wild-type key genomic features. The collection sites of BANAL-20-52 and BANAL-20-236 were localized in caves named Pha Nouk Kok (18.53N, 101.98E) and Pha Tong (18.55N, 101.98E) with a short distance of 2.22 km. In addition, both of BANAL-20-52 and BANAL-20-236 had been detected in the samples, isolated from malayan horseshoe bats (Rhinolophus malayanus).

The origin of SARS-CoV-2 in bats

The data of the three close relatives were sourced from three independent studies [4, 11, 12], respectively. These studies contributed a total of complete or partial genomic sequences from 156 bat Sarbecovirus strains, which were compiled into a dataset containing a wealth of geographic information for our further analysis (Materials and Methods). Among the 156 bat Sarbecovirus strains, 111 were used in the previous study [11]. Based on a phylogeographic analysis, the previous study [11] proposed a divergent evolution between sarbecoviruses from subtropical northern Vietnam and those from tropical southern Vietnam, emergence of SARS-CoV in horseshoe bats from northern Yunnan and emergence of SARS-CoV-2 in horseshoe bats from southern Yunnan. Although the phylogeographic analysis applied a new method, named coloured genomic bootstrap (CGB) barcode, it still exhibited two limitations. The first pertains to the absence of S18CXBatR24 in the analyzed dataset. This omission renders misleading information that none bat Sarbecovirus strains from Yunnan of China harbor wild-type RC6 and RC7. The second limitation lies in the inherent nature of phylogeographic analysis, which can only provide estimates of the relative relationships between different strains, requiring concrete evidence, like wild-type RC1, to determine the directionality of these relationships. Indeed, using phylogeographic analysis for the origin tracing of SARS-CoV-2 does not yield additional information beyond what is obtained from the simple analyses based on nucleotide identities.

Different from the previous study [11], we used the geographical information associated with the 156 bat Sarbecovirus strains to delineate the distribution of the seven wild-type key genomic features across Southeast Asia and South China (Fig 3). Generally speaking, most of the bat Sarbecovirus strains did not harbor more than three wild-type key genomic features. Notably, only the three close relatives and RaTG13 harbored seven and five wild-type key genomic features, respectively. These new findings were consistent with our previous conclusion that the genesis of SARS-CoV-2 was determined by several key or crucial genomic features, integrated through several key or crucial recombination events. Wild-type RC6 and RC7 are not rare among bat Sarbecovirus, suggesting that they might not have made crucial contributions to the genesis of SARS-CoV-2. However, possessing wild-type RBDs, equivalent to harboring wild-type RC6 and RC7, is a basic requirement for bat Sarbecovirus strains to evolve into SARS-CoV-2. In contrast, wild-type RC0 and RC1 are rare, suggesting they may have made crucial contributions to the genesis of SARS-CoV-2. As wild-type RC1 was only detected in a bat Sarbecovirus population in Laos and wild-type RC0 remained undetected in bats, wild-type RC0 and RC1 may have been integrated into SARS-CoV-2 through the last and second-to-last type 2 recombination events, respectively. In addition, wild-type RC0 in SARS-CoV-2 may have been acquired in intermediate hosts or humans [5].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The geographical distribution of studied Sarbecovirus strains.

Sample collection site 1–19 and 20–23 were reported in the previous study [4, 11], respectively. Site 4, 6, 7, 9, 11, 14, 15, 16, 17 and 19 were not shown in this figure, as no Sarbecovirus were detected in these locations. Three close relatives (BANAL-20-52, Rp22DB159, and S18CXBatR24) were discovered in Feuang, Vientiane province of Laos (Feuang), Dien Bien municipality of Vietnam (DB-3), and Chuxiong, Yunnan province of China (CX), respectively, while RaTG13 were discovered in Tongguan town, Yunnan province of China (TG). The distances between CX (25.22 N,101.38 E) and the outbreak site (30.61 N, 114.24 E) is 1,397 km (Not shown); the distance between TG (23.14 N, 101.23 E) and CX is 232 km; the distance between Feuang (18.53 N, 101.98 E) and CX is 746 km; the distance between DB-3 (22.39 N, 102.24 E) and CX is 327 km. The distances of Feuang-TG, Feuang-DB-3, DB-3-TG are 518, 430 and 133 km, respectively. Wild-type RC1 was only detected at site 20. We proposed that SARS-CoV-2 originated along a conceptual path from Yunnan, China to Vietnam, and then to Laos, acquiring crucial genomic features along the way. 3, 23, 26 and 31 are DB-3, Feuang, CX and TG. The latitudes and longitudes of 32 sample collection sites are provided in S1 Text.

https://doi.org/10.1371/journal.pone.0309557.g003

Despite wild-type RC0 not being detected outside of the SARS-CoV-2 genome, integration of the new crucial evidence (wild-type RC1) connected previously sparse and isolated evidence into a coherent chain, concentrating on the key genomic features and the three close relatives from Feuang, Vientiane province of Laos (Feuang), Dien Bien municipality of Vietnam (DB-3), and Chuxiong, Yunnan province of China (CX). It’s highly improbable that these three close relatives independently evolved to harbor seven wild-type features as SARS-CoV-2. Instead, the origins of the three close relatives involved Sarbecovirus populations from Feuang, DB-3, and CX (Fig 3), suggesting a complex pattern of viral transmission and recombination events across various locations. Since wild-type RC1 was only detected in the Sarbecovirus population in Feuang, a putative direction of the SARS-CoV-2 genesis could be inferred from either DB-3 or CX to Feuang, representing Vietnam or Yunnan of China, to Laos (Fig 3). The genesis of SARS-CoV-2 likely involved a few key recombination events in limited Sarbecovirus groups within a limited number of regions. These key recombination events could have occurred in three bat hosts: malayan horseshoe bats (Rhinolophus malayanus), least horseshoe bats (Rhinolophus pusillu), and marshall’s horseshoe bats (Rhinolophus marshalli). The last or second-to-last recombination events crucial to the genesis of SARS-CoV-2 could have occurred in malayan horseshoe bats in tropical regions, particularly Feuang or its vicinity.

The direction of the SARS-CoV-2 genesis between DB-3 and CX can not be determined by key genomic features, given that both Rp22DB159 from DB-3 and S18CXBatR24 from CX harbor seven wild-type features. Furthermore, Rp22DB159 and S18CXBatR24 share a nucleotide identity of 97.38%, which is the highest among all identities between the 156 strains. According to a previous study, Sarbecovirus genomes from lower latitudes have a stronger synonymous nucleotide composition (SNC) bias in favour of U nucleotides than those from higher latitudes. While, our study revealed an increase in A-content in Sarbecovirus genomes of S18CXBatR24, Rp22DB159, BANAL-20-52 from 29.7, 29.78 to 30.32%, suggesting the direction of the genesis of SARS-CoV-2 from CX to DB-3. Thus, we proposed that SARS-CoV-2 originated along a conceptual path from Yunnan, China to Vietnam, and then to Laos (Fig 3), acquiring crucial genomic features through crucial recombination events. However, it remains uncertain whether the driving force behind this path is attributable to climate or latitude, given that the three regions (CX, DB-3, and Feuang) span from subtropical to tropical climates, with latitudes ranging from 2000 m, 900 m to 3~400 m, respectively. Furthermore, the underlying mechanisams, particularly at the molecular level remain unknown. Notably, the increase in A-content has the potential to augment the NSP15 cleavage sites within the genome, possibly leading to an increased frequency of recombination events. According to our previous study [8, 9], the NSP15 cleavage sites contain U nucleotides at 5’ ends, corresponding to A nucleotides in the TRS-like sequences and the frequency of recombination events dependents on the cleavage of synthesized RNAs at the NSP15 cleavage sites. Additionally, the similar A-content observed in SARS-CoV-2 and BANAL-20-52 (30.31% vs. 30.32%) could indicate shared evolutionary pressures or a common genetic background. In contrast, RaTG13 has a lower A-content 29.94%, although it has the similar nt identities with SARS-CoV-2 as BANAL-20-52 (96.13% vs. 96.85%).